If you search the forum you'll see that others have come across this issue. The secret is:
There are other good suggestions in that thread too.
Also, you'll likely only need to enter in the forward primer, i.e.--p-front, often you'll not need to enter in the adapter sequence unless you expect read-through.
I tried it this way and it worked fine. Though I can still see many of sequences with primer in the front. Not sure why it's not working. Attached is the snapshot of the rep seq table.
I would not recommend doing this as you will "see" your primers in the output. The problem is that you are not allowing for any error within the span of the primer sequence. Unless you allow for some errors / mismatches / indels (default is 10%), they will not be trimmed from the sequence as you are requiring the match to be exact. For example, the second sequence has a 1 bp mismatch to the primer. Sequence 5 and 6 are missing a base
Note: the M-A is not a mismatch with --p-match-adapter-wildcards set, as M represents A or C.
Also, I had mentioned there are other tricks in that thread I linked. For example you can set --p-discard-untrimmed as outlined in that very same thread, here.
So leave --p-error-rate as default and set --p-discard-untrimmed.